By Kate Robards, senior writer
Physicians spend a substantial amount of time on documentation, often more than they spend with their patients. An article published in JAMA Network Open found that primary care visits average about 30 minutes, while physicians typically spend 36 minutes interacting with the electronic health record (EHR) per visit. The burden of documentation is a well-known challenge across medical specialties, and sleep medicine is no exception.
Artificial intelligence (AI) is emerging as a tool to help lighten that load. In sleep medicine, AI applications already span a broad range, from scoring sleep studies and assisting with CPAP mask fittings to helping draft marketing materials and review legal contracts. AI tools can even help identify cybersecurity vulnerabilities, assist with clinical research and draft educational materials. One increasingly common application is the use of AI scribes to generate clinical documentation.
AI scribes rely on large language models, which are trained to understand and generate human-like text. These tools can record patient-provider conversations and produce draft documentation that summarizes the encounter, often including assessments and treatment plans. Some also support dictation, allowing clinicians to speak their notes directly into the system.
The primary appeal of AI scribes is their ability to reduce documentation time. By generating notes during or after a patient visit, these tools can allow clinicians to focus more on their interaction with the patient rather than entering information into the computer. Some tools also assist with preparing charts before a visit, summarizing the discussion for the patient or formatting after-visit summaries.
Despite these advantages, limitations remain. Some AI scribes struggle with the nuanced terminology of specialties like sleep medicine. Because many tools are developed with primary care in mind, their templates may not align with the needs of sleep practitioners. As a result, clinicians often need to spend additional time reviewing and editing AI-generated notes to ensure accuracy. There is also a learning curve involved in integrating these tools into clinical workflows, and perfectionists may find the occasional omissions or reordering of information frustrating.
Privacy, security, and compliance are other important considerations. Not all AI tools are HIPAA-compliant, and using noncompliant software could present legal risks. As with any clinical documentation tool, final responsibility for the content lies with the health care professional, so maintaining a careful review process is essential.
Some AI scribes are integrated into widely used EHR systems such as Epic, Cerner and Athena. These tools are being used by some sleep medicine professionals, particularly in academic or multi-specialty settings. While most AI scribe platforms are not specifically marketed for sleep practices, several can be adapted to support common documentation needs in the field. Systems that allow customization or support specialty templates can be adjusted to align with sleep workflows. For example, tools like DAX, Abridge and DeepScribe are frequently used in specialties where detailed patient interviews are central, which is also the case in sleep consultations. Others, such as Freed, offer documentation using the SOAP format (subjective, objective, assessment and plan), which can be useful for recording subjective sleep complaints, test results and treatment plans.
Before adopting an AI scribe, several practical and ethical questions should be addressed. Patient consent and transparency are essential, particularly for tools that involve audio recordings. Organizations should ensure that tools are HIPAA-compliant and approved for use within their systems. In private practices, selecting secure tools and developing internal policies can help reduce risk.
Transparency in documentation is also recommended. The American Medical Association recommends that when AI is used to generate patient-facing content, such as care summaries, its use should be disclosed or clearly noted within the content. This promotes clarity and helps support communication with patients. It is also important to be aware that AI systems may reflect biases present in their training data, which could affect the way clinical notes are generated. Regular review of documentation helps mitigate this risk.
As interest in these tools grows, sleep professionals have access to a variety of resources. The American Academy of Sleep Medicine offers guidance through its AI in Sleep Medicine Committee, which provides webinars and educational content. AASM Link, a new AI-powered platform, includes a searchable resource library that helps sleep professionals quickly access clinical practice guidelines, standards of practice and essential references.
AI scribes are not a one-size-fits-all solution, and they do not eliminate the need for clinical judgment or oversight. However, when implemented thoughtfully, they may help reduce administrative tasks and make clinical workflows more efficient. For sleep professionals considering these tools, a careful approach that emphasizes security, accuracy, and patient transparency will be key.
